Every vendor deck this year has a slide with three eras on it, and the third era is always the one they are selling. The slide is usually right about the shape and useless about the question that matters to an operator, which is: what era is my company actually in, and what would it take to move?
This is that question, answered for a company of 50 to 500 people running a CRM, a set of books, something that counts inventory or hours, and a support desk. The eras are real. They are defined by what the system is allowed to do with a record, not by what the licence says.
The first era is a database with a form on the front. A person does something in the world, then a person types what happened into the system. The CRM holds the deal because a rep updated the stage. The books hold the invoice because someone raised it. The warehouse count is right if the last person who touched the shelf also touched the screen.
The tell for era one is simple. Ask what happens if nobody logs in for a day. In an era-one estate, nothing. The business carries on, the systems fall behind, and someone catches up the records on Friday. The system is a memory, and a leaky one.
Most companies would say they left this era years ago. Most of their systems have not. The accounting package is almost always era one by design, and that is fine, because a ledger should record what happened rather than decide it. The problem is the CRM that is era one by neglect, where the pipeline is whatever the reps remembered to type.
The second era adds rules. A record changes and the system moves work to the next person. A lead comes in and an assignment rule gives it an owner. A deal reaches a stage and a task appears for the person who has to send the quote. An invoice goes past due and a reminder goes out. Workflow automation, approval chains, notifications, the Blueprint in Zoho CRM, the automations in every help desk.
The tell for era two: the system knows the next step and makes a person take it. It does not take the step. It is a very good dispatcher. A great deal of operational improvement in the last decade was moving from era one to era two, and a company that has done it properly feels it: fewer things fall between people, and the handoffs are visible.
Era two has a ceiling, and the ceiling is the person at the end of every rule. The rule can route the quote request to the rep. It cannot write the quote, because writing the quote requires reading the account history, the price list, the stock position and the last three emails, and deciding. We wrote about where that line sits in the piece on agentic automation versus workflow automation: the difference is who decides.
The third era is the one on the slide. The system reads the record, reads the records around it in other systems, decides within rules that a person set, changes the record, and leaves a trail. The agent drafts the quote from the account history and the price list and presents it. The agent reconciles the three inventory counts and flags the one that disagrees, with the reason. The agent reads the support ticket, checks the contract, and proposes the resolution the contract allows.
The tell for era three is the one most companies get wrong. It is not that the system has a language model in it. A chat window bolted onto a CRM is era one with a nicer form. The tell is that the system changes records on its own authority, inside a boundary a person drew, and a person can see every change and why. We described the shape of that in what changes when the system acts on the record: the money is in the agent that works across three systems, and so is the risk.
The agentic enterprise, in the vendor sense, is a company whose systems are mostly era three. Almost no company of 50 to 500 people is there, and the honest ones know it.
Here is the part the slide leaves out. A company does not move eras by buying era-three software. It moves eras when its records can carry the weight of the era.
Era one needs records that exist. Era two needs records that are complete enough for a rule to fire on, which is why so many workflow rules are switched off six months after launch: the field the rule depends on is blank on a third of the records. Era three needs records that are complete, consistent across systems and owned, because an agent that acts on a wrong record acts wrongly with confidence, and nobody is at the end of the rule to catch it.
So the test for your estate is a test of records. For each system, ask three questions. Does every record the system holds exist in it, or do some live in email and spreadsheets? Are the fields a rule would need filled on nearly every record, or on the ones someone cared about? When two systems hold the same thing, the customer, the product, the price, do they agree, and is there a rule for which one wins? We wrote the write rule up in system of record versus source of truth: exactly one system may write each business fact.
A system that fails the first question is era one, whatever the licence. A system that passes the first and fails the second is era one with era-two features turned off. A system that passes the first two and fails the third can run era-two automation inside itself and cannot safely run an era-three agent across its neighbours, because the agent will act on the disagreement.
Run the test and a familiar picture appears. The books are era one and should stay there. The CRM is era two inside its own walls, with a dozen rules routing leads and tasks, and era one at the edges, where the account record disagrees with the books about who the customer is. The warehouse or project system is era one with an export. Support is era two. Nothing is era three, and the licence for era three has been paid for since the renewal.
That is not a failure. It is the normal state of a company that grew faster than its systems, and it is the state most agent projects start from, which is why so many of them stall. The agent is asked to act across records that do not agree, and either it acts and makes a mess, or it is wrapped in so many human approvals that it is an era-two rule with a larger invoice.
The move from era two to era three is, in order, a records project, a permissions project and then a software project, and the first two are most of the work.
The records project is making the three questions above come out yes for the systems the agent will touch, starting with the customer, because every agent reads the customer. The permissions project is deciding what the agent may change without asking, what it must propose, and what it may not see, and writing that down where the system enforces it rather than where a slide describes it. The software project is the smallest part: the agent, the connections, the trail.
Done in that order, the agent arrives into an estate that can carry it, and the first thing it does is useful. Done in the other order, which is the order the licence encourages, the agent arrives first and spends its first quarter discovering the records.
This is the work we do as a managed subscription rather than a project, because the shift is not a cutover with a date. The records drift, the rules change as the business does, and the agent's boundary moves as trust is earned. One accountable party runs the three layers and is measured on whether the agent's actions were right, not on whether the software was installed.
If you want to know which era each of your systems is in, we score the three questions against your estate in a no-risk discovery, and you pay only if you go ahead.
In a no-risk discovery we look at your CRM, finance and operations systems and who owns each part, and show what a connected system would do differently. You pay only if you proceed. Or see how we approach it.
More on the same problem:
"Something isn't working." "We're outgrowing our tools." "We're afraid to make the wrong move." No-Risk Discovery is a short, practical conversation that gets you clarity before you commit to anything big. We'll tell you if we're a fit. If we're not, we'll tell you that too.
Say who you are and ask. A person replies within one business day.
